Related Experiment Video
Updated: Jun 14, 2026
13:44
Detection of Architectural Distortion in Prior Mammograms via Analysis of Oriented Patterns
Published on: August 30, 2013
43.0K
Combined use of two artificial intelligence-based algorithms for mammography triaging: a retrospective simulation
Hee Jeong Kim1, Hak Hee Kim2, Hye Joung Eom1
1Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, South Korea.
European Radiology
|August 21, 2025
Summary
Combining two artificial intelligence (AI) algorithms for mammogram interpretation improved accuracy and reduced the number of mammograms needing radiologist review. This AI approach enhances workflow efficiency in breast cancer screening.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Healthcare
- Breast Cancer Diagnostics
Background:
- Mammography interpretation is crucial for early breast cancer detection.
- Radiologist workload can be high, potentially impacting interpretation efficiency.
- Artificial intelligence (AI) algorithms show promise in assisting with mammogram analysis.
Purpose of the Study:
- To evaluate the effectiveness of combining two commercial AI algorithms for mammogram interpretation.
- To assess different triaging scenarios using AI to enhance accuracy and reduce radiologist workload.
- To compare the performance of combined AI algorithms against single AI and manual interpretation.
Main Methods:
- 3012 mammograms (screening/diagnostic) with 213 cancer cases were analyzed.
- Two AI algorithms (AI-1, AI-2) categorized mammograms by malignancy risk.
- Five triaging scenarios (Sensitive, Specific, Conservative, Sequential A/B) were tested to determine recall, negative classification, or radiologist review.
Main Results:
- Sensitive Mode: 84% sensitivity, outperforming single AI and manual reading, with an 18.3% reduction in mammograms for review.
- Specific Mode: 87.7% specificity, exceeding single AI and comparable to manual reading, with a 41.7% reduction in mammograms for review.
- Other modes showed comparable sensitivity to single AI/manual reading, with significant reductions (up to 49.8%) in mammograms requiring review.
Conclusions:
- Combining two AI algorithms enhances mammography interpretation by improving sensitivity or specificity.
- AI-driven triaging significantly reduces the number of mammograms requiring radiologist review.
- Scenario selection is key and requires validation in broader screening populations.
